Conservation of Aging and Cancer Epigenetic Signatures across Human and Mouse ul F. Perez,†,1,2,3,4,5 Juan Ram Ra on Tejedor,†,1,2,3,4,5 Pablo Santamarina-Ojeda,1,2,3,4,5 1,2,3,4,5 Virginia Lopez Martınez, Rocıo G. Urdinguio,1,2,3,4,5 Lucıa Villama~ nan,6 Ana Paula Candiota,6,7 N mı Vidal Sarr o,8 Marta Barradas,9 Pablo Jose Fernandez-Marcos,9 Manuel Serrano,10,11,12 Agusın F. Fernandez,*,1,2,3,4,5 and Mario F. Fraga*,1,2,3,4,5 1 Cancer Epigenetics and Nanomedicine Laboratory, Nanomaterials and Nanotechnology Research Center (CINN-CSIC), University of Oviedo, Oviedo, Spain 2 Institute of Oncology of Asturias (IUOPA), University of Oviedo, Oviedo, Spain 3 Health Research Institute of Asturias (ISPA), University of Oviedo, Oviedo, Spain 4 Department of Organisms and Systems Biology (B.O.S.), University of Oviedo, Oviedo, Spain 5 Rare Diseases CIBER (CIBERER) of the Carlos III Health Institute (ISCIII), Madrid, Spain 6 Unitat de Bioquımica de Biociències, Departament de Bioquımica i Biologia Molecular, Edifici Cs, Universitat Autonoma de Barcelona, Cerdanyola del Vallès, Spain 7 Centro de Investigacion Biomedica en Red en Bioingenierıa, Biomateriales y Nanomedicina (CIBER-BBN), Madrid, Spain 8 Servicio Anatomıa Patologica, Hospital Universitari de Bellvitge—IDIBELL, Hospitalet de Llobregat, Spain 9 Metabolic Syndrome Group—BIOPROMET, Madrid Institute for Advanced Studies—IMDEA Food, CEI UAMþCSIC, Madrid, Spain 10 Tumour Suppression Group, Spanish National Cancer Research Centre (CNIO), Madrid, Spain 11 Cellular Plasticity and Disease Group, Institute for Research in Biomedicine (IRB Barcelona), Barcelona Institute of Science and Technology (BIST), Barcelona, Spain 12 Catalan Institution for Research and Advanced Studies (ICREA), Barcelona, Spain † These authors contributed equally to this work. *Corresponding authors: E-mails: agustin.fernandez@cinn.es; mffraga@cinn.es. Associate editor: Li Liu Abstract Aging and cancer are two interrelated processes, with aging being a major risk factor for the development of cancer. Parallel epigenetic alterations have been described for both, although differences, especially within the DNA hypome- thylation scenario, have also been recently reported. Although many of these observations arise from the use of mouse Article models, there is a lack of systematic comparisons of human and mouse epigenetic patterns in the context of disease. However, such comparisons are significant as they allow to establish the extent to which some of the observed similarities or differences arise from pre-existing species-specific epigenetic traits. Here, we have used reduced representation bisulfite sequencing to profile the brain methylomes of young and old, tumoral and nontumoral brain samples from human and mouse. We first characterized the baseline epigenomic patterns of the species and subsequently focused on the DNA methylation alterations associated with cancer and aging. Next, we described the functional genomic and epigenomic context associated with the alterations, and finally, we integrated our data to study interspecies DNA methylation levels at orthologous CpG sites. Globally, we found considerable differences between the characteristics of DNA methylation alterations in cancer and aging in both species. Moreover, we describe robust evidence for the conservation of the specific cancer and aging epigenomic signatures in human and mouse. Our observations point toward the preservation of the functional consequences of these alterations at multiple levels of genomic regulation. Finally, our analyses reveal a role for the genomic context in explaining disease- and species-specific epigenetic traits. Key words: epigenetics, DNA methylation, histone modification, mouse, human, conservation. Introduction consequently analogous epigenetic alterations have been Aging is one of the main risk factors associated with the identified for both, providing explanations for the possible development of cancer. These two processes have been ex- molecular links between them (Aunan et al. 2017; Yu et al. tensively investigated from the epigenetic perspective and 2020). These alterations mainly relate to local gains in DNA ß The Author(s) 2021. Published by Oxford University Press on behalf of the Society for Molecular Biology and Evolution. This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com Open Access Mol. Biol. Evol. 38(8):3415–3435 doi:10.1093/molbev/msab112 Advance Access publication April 19, 2021 3415 Perez et al. . doi:10.1093/molbev/msab112 MBE methylation at CpG-dense regions and global losses at the dynamics associated with cancer and aging in both human genomic scale involving repetitive DNA (Baylin and Jones and mouse. To this end, we profiled the brain methylomes of 2016; Sen et al. 2016). Nonetheless, the extent of these paral- young and old, tumoral and nontumoral brain samples from lelisms has not as yet been clearly outlined, particularly within human and mouse by reduced representation bisulfite se- the hypomethylation scenario (Dmitrijeva et al. 2018; Perez quencing (RRBS), which provided a robust and interspecies- et al. 2018). Moreover, recent whole-genome studies in comparable platform for the analysis at hand (Bock et al. mouse have also failed to confirm hallmarks such as global 2010). We first characterized the baseline epigenomic pat- hypomethylation with aging in tissues including liver, hippo- terns of both species and subsequently focused on the campus, and stem cells (Sun et al. 2014; Cole et al. 2017; Hahn DNA methylation alterations associated with cancer and ag- et al. 2017; Masser et al. 2017; Hadad et al. 2019; Hernando- ing. Next, we described the functional genomic and epige- Herraez et al. 2019). These observations underscore two dif- nomic context associated with the alterations, and finally, we ferent, but equally important, issues: first, there is a need to integrated our data to study interspecies DNA methylation study aging- and cancer-associated epigenetic patterns in a levels at orthologous CpG sites. Globally, we found consider- systematic manner to facilitate an integrated comparison be- able differences between the genomic and epigenomic char- tween the two processes; second, it is essential to also confirm acteristics of DNA methylation alterations in cancer and whether aging- and cancer-associated epigenetic alterations aging in human and mouse. Furthermore, we describe robust are equivalent in human and mouse models, as the latter are evidence for the conservation of the specific cancer- very important experimental proxies of human biology. and aging-associated epigenomic patterns in both species. Mouse models have been widely used in the characteriza- Our observations point toward the preservation of the func- tion of landmark epigenetic mechanisms such as transgenera- tional consequences of these alterations at multiple levels of tional inheritance and genomic imprinting (Blewitt and genomic regulation, including the locations and chromatin Whitelaw 2013) as well as developmental epigenomic reprog- context of the alterations, the genetic pathways, and tran- ramming (Hanna et al. 2018), and they are often used to scription factors (TFs) involved. Additionally, the analysis characterize alterations in specific disease-associated path- of orthologous CpG loci allowed us to describe species- ways which sometimes have a translational correspondence common locations of DNA methylation variability and in human (Espada and Esteller 2013). To date, reports suggest species-common cancer- and aging DNA methylation alter- that DNA methylation patterns preserve global features such ations. Finally, our analyses suggest that the genomic context as the presence of nonmethylated islands and gene body associated with many CpG sites may help explain the obser- methylation across mammalian species in general (Long vation of species-discordant DNA methylation as well as can- et al. 2013; Schroeder et al. 2015), with other parallelisms cer and aging-specific DNA methylation alterations. extending across more distant vertebrate taxa (Elango and Yi 2008). More directed comparisons involving mouse and Results and Discussion human have described a general conservation of their genome-wide epigenetic patterns (Edwards et al. 2010) and Baseline Epigenomic Patterns Are Conserved between have particularly focused on the interspecies conservation of Human and Mouse tissue-specific methylation patterns (Kessler et al. 2016; Zhou To characterize the intricate, conserved epigenetic signatures et al. 2017; Chen et al. 2018) and developmental processes that may exist in aging and cancer, we studied the genome- (Lister et al. 2013). Nonetheless, considerable epigenetic differ- wide DNA methylation patterns of young, old, tumoral, and ences in proximal species such as primates have been nontumoral brain samples from human and mouse (n ¼ 24, 3 reported (Hernando-Herraez, Heyn, et al. 2015; Mendizabal per group, fig. 1A). We produced methylation values for an et al. 2016). So far there have been only limited high- average of 3.94 million and 1.62 million unique CpG sites per resolution studies that focus on directly assessing specific sample for human and mouse, respectively, and retained epigenetic similarities between human and mouse in either a final total of 1,171,918 and 585,234 CpG sites common aging (Maegawa et al. 2017; Wang et al. 2017) or cancer to all samples after filtering for coverage (see Materials and (Maegawa et al. 2014), and those that exist are often limited Methods and supplementary table 1, Supplementary Material by their compilation of data from different studies or the use online, for sample and alignment information). We first of different profiling technologies for each species. Some of sought to characterize the global aspects of DNA methylation these issues have arisen from the lack of a large-scale DNA and its relationship with genomic function in the two species methylation array technology for mouse. Overall, then, there in order to determine the baseline similarities and differences is a lack of back-to-back human–mouse comparisons of DNA in DNA methylation patterns in our study system. For this methylation at single-base resolution in the context of dis- initial analysis, we focused on the nontumoral samples be- ease. Even so, this issue is particularly relevant in the study of cause the tumors showed extensive deviation from the rest of DNA methylation alterations found independently in both the samples using pairwise Pearson correlations (fig. 1B), and species in order to reliably ascertain the extent to which some our goal was to profile the general species-wise features of of the observed similarities or differences arise from pre- DNA methylation in our samples. existing species-specific epigenetic traits. Though significant, we did not observe noticeable differ- In this study, we sought to perform a systematic, inte- ences between the species in the distribution of the overall grated, and intercomparable analysis of the DNA methylation methylation values, with an average methylation of 0.266 for 3416 Aging and Cancer Epigenetic Signatures across Human and Mouse . doi:10.1093/molbev/msab112 MBE A B C D E F G H I J FIG. 1. Baseline epigenomic patterns in human and mouse. (A) Schematic of the study design. (B) Heatmaps showing the pairwise Pearson correlations between DNA methylation levels of the profiled CpG sites across samples. (C) Violin plots of global levels of DNA methylation for the pooled samples (red dots indicate average values; h, human; m, mouse; tumor samples are excluded). (D and E) Barplots indicating the distribution of the profiled CpG sites in human and mouse across CpG island and gene-region locations. (F and G) Violin plots showing the DNA methylation levels for the pooled samples with the CpG sites grouped by CpG island locations and gene-region locations (red dots indicate average values). (H) Density plots indicating the relationship between DNA methylation values and surrounding density (2-kb bins) in CpG sites for the profiled CpGs. (I and J) Heatmaps and bubble plots showing the significant overenrichments in log2(odds ratio) of CpG sites classified as high-, medium-, and low- methylation in ChIP-seq peak locations of different histone marks across a panel of tissues or in the locations of repetitive DNA classes across the human and mouse genomes. 3417 Perez et al. . doi:10.1093/molbev/msab112 MBE human and 0.272 for mouse (fig. 1C, Wilcoxon rank-sum better study the relationship of DNA methylation concerning P ¼ 0.02). The low levels of methylation observed are to be its genomic function in mouse and human, we integrated our expected since the RRBS technology targets CpG-dense data with ENCODE and NIH Roadmap Epigenomics ChIP-seq regions, such as CpG islands, which tend to be unmethylated. data profiling the histone modifications H3K4me1, H3K4me3, On a genome-wide scale, however, most CpG sites are meth- H3K27ac, H3K36me3, H3K27me3, and H3K9me3. We focused ylated in mammals (Greenberg and Bourc’his 2019), and tech- on a panel of tissues which was available for the two species, nologies such as whole-genome bisulfite sequencing (WGBS), and contained several brain tissue tracks, analyzing a com- which also targets intergenic regions, would detect higher bined total of 144 data sets. We classified CpG sites as high-, levels of DNA methylation. medium-, or low-methylation loci (>0.8, [0.2,0.8], and <0.2 In order to ascertain that the RRBS was evaluating a com- average methylation values, respectively) and investigated the parable population of CpGs for each species, we annotated colocalization of these sites with the histone marks by over- the sites to functional genomic locations in terms of their enrichment analyses at FDR <0.05 (see Materials and CpG island membership status and genetic elements. For Methods). With this procedure, we observed practically iden- both species, the technology interrogated a majority of tical chromatin signatures of methylation status in human CpG sites located at CpG islands (fig. 1D), with a slight in- and mouse (fig. 1I and supplementary table 2, Supplementary crease in open sea CpGs for mouse (Fisher’s P < 0.001, Material online): high methylation sites were associated with OR ¼ 1.89). Regarding genetic elements, the distribution of active gene-body H3K36me3 and repressive H3K9me3 marks, the quantified CpGs was also very similar for both species intermediate methylation also appeared to be related to (fig. 1E), with the largest proportion of sites being located at enhancer/promoter-associated H3Kme1 and low methyla- promoters and gene bodies. These observations suggest that, tion sites were strongly associated with the active promoter for human and mouse, the RRBS technology targets popula- and enhancer H3K4me3 and H3K27ac marks. These observa- tions of CpG sites which have analogous genomic contexts. tions confirm that the relationship between DNA methyla- Next, we proceeded to characterize the relationship be- tion and genomic function is conserved in both species tween DNA methylation and CpG context in both human (Roadmap Epigenomics Consortium et al. 2015; Gorkin and mouse. With respect to CpG islands, DNA methylation et al. 2020). was highest at open sea locations and lowest at islands for One of the genomic elements to which DNA methylation both species, with a transition in methylation across shelves has been most functionally linked is repetitive DNA (Beisel and shores (fig. 1F). For gene elements, the relationship with and Paro 2011). Hence, we mapped the classified CpG sites to DNA methylation was also parallel between human and RepeatMasker repetitive elements and looked for overenrich- mouse, with promoters and 50 -UTR regions showing the low- ment in specific repeat classes. Looking at the most common est values, followed by gene body exons and introns, whereas repeats, we observed, for both human and mouse, that high- intergenic regions were those with the highest DNA methyl- methylation loci were associated with SINE, LINE, LTR, and ation (fig. 1G). The presence of DNA methylation is inversely DNA repeats, whereas low-methylation was linked to simple correlated to the density in surrounding CpG sites (Chen et al. and low-complexity repeats (fig. 1J and supplementary table 2018), and this may help explain some of the patterns found 3, Supplementary Material online). Expanding the analyses to at genomic locations, especially concerning CpG islands and repetitive element families also revealed parallel patterns in open sea. We therefore characterized the density in the CpG the two species (supplementary fig. S2, Supplementary sites associated with each locus by counting the number of Material online). These results are also in consonance with neighboring CpG sites in surrounding 2-kb windows. As existing literature both for mouse and human (Edwards et al. expected, in general, low methylation values were found as- 2010). sociated with locations with higher density in CpG sites, whereas higher methylation values were found at low-density Dynamics of DNA Methylation Alterations in Human locations, for both species (fig. 1H). Indeed, the different ge- Aging and Cancer nomic locations (islands and gene regions) displayed different Once we had confirmed the analogous human and mouse distributions of density in CpG sites (supplementary fig. S1A, baseline epigenomic patterns in our study system, we pro- Supplementary Material online) and, moreover, the mean ceeded to determine the DNA methylation alterations asso- methylation value of each element correlated almost per- ciated with aging and with cancer for each species fectly with its average density in CpG sites (supplementary independently. Starting with human, we found no general fig. S1B, Supplementary Material online). Collectively, these DNA methylation differences associated with cancer or aging results confirm that the relationship between DNA methyla- by looking at all the sites analyzed by the RRBS, although tion and functional genomic elements such as CpG islands tumor samples tended to have higher methylation values and gene regions is conserved between human and mouse (fig. 2A). Moreover, PCA clustering suggested differences be- and is mostly explained by the density of CpG sites associated tween the experimental groups as they segregated across the with each element (Edwards et al. 2010). most meaningful dimensions (fig. 2B). We performed a dif- In the regulation of gene expression, epigenetic marks such ferential methylation analysis with a stringent threshold (FDR as DNA methylation are not independent, rather, they act in <0.05, average change in methylation >0.2) to define differ- combination with interrelated marks such as histone mod- entially methylated CpGs (dmCpGs) in aging and cancer. We ifications to define functional states (Du et al. 2015). Thus, to identified widespread alterations in both processes, counting 3418 Aging and Cancer Epigenetic Signatures across Human and Mouse . doi:10.1093/molbev/msab112 MBE A B C D E F G H FIG. 2. Cancer and aging DNA methylation alterations in human. (A) Violin plots of global levels of DNA methylation of all CpGs across the profiled samples (N, nontumor; T, tumor; Y, young; O, old). (B) PCA plots showing the distribution of the samples across the first two principal components, with the percentage of variance each explains. (C) Heatmaps showing the methylation values of the top 1,000 dmCpGs found in the aging and cancer comparisons. (D) Barplots indicating the relative numbers of hyper- and hypo-dmCpGs (differentially methylated CpGs) in aging and cancer. (E and F) Barplots indicating the distribution of the dmCpGs across CpG island and gene-region locations, including the distribution of the universe of CpGs profiled by the RRBS (“all”). (G) Lollipop plots comparing the number of dmCpGs mapped to genes (left vertical axis, black dots) with the number of different genes that the dmCpGs mapped to (right vertical axis, orange dots). (H) UpSet plot describing the dmCpG sets and their intersections, indicating the size of the sets and the fold-enrichment (FE) of the intersections based on their expected overlaps. 154,770 dmCpGs associated with cancer and 66,838 with ag- comparison using all nine nontumoral samples (nontumor, ing (lists of dmCpGs are available as extended data sets), with young, and old groups) as the control group yielded compa- hypermethylation dominating the cancer scenario particu- rable results and numbers of dmCpGs (supplementary fig. larly (accounting for 72% of changes, fig. 2C and D). S3A, Supplementary Material online), indicating that the Furthermore, performing the cancer differential methylation cancer-associated alterations observed are robust and are 3419 Perez et al. . doi:10.1093/molbev/msab112 MBE independent of the samples used. When mapping the performed pathway enrichments on these intersections and dmCpGs to CpG island locations (fig. 2E), we observed that only observed enrichments in common CpGs between can- hypermethylation occurred for cancer preferentially at CpG cer hypermethylation and aging hyper- and hypomethylation islands, with a distribution akin to that of the RRBS back- (full results are available as extended data sets). These results ground of analyzed sites, whereas hypomethylation and suggest that the functional commonalities between DNA aging-associated changes were, in general, enriched at open methylation deregulation in human cancer and aging mainly sea locations (Fisher’s P < 0.001, ORs ¼ 8.9, 3.3, and 4.0, pertain to Polycomb-target gene hypermethylation in cancer, respectively). Regarding gene locations, cancer hypomethyla- whereas these pathways manifest both hyper- and hypome- tion and aging hyper- and hypomethylation were particularly thylation in aging. increased at intergenic- and intronic sequences (Fisher’s P < 0.001, intergenic ORs ¼ 4.5, 2.0, and 2.0, intronic Dynamics of DNA Methylation Alterations in Mouse ORs ¼ 1.8, 1.7, and 2.0). In line with this observation, cancer Aging and Cancer hypermethylation occurred more frequently at CpG-denser Next, we profiled the DNA methylation alterations associated regions than the rest of the changes (supplementary fig. S3B, with aging and cancer in mouse. First, regarding the general Supplementary Material online) and, additionally, it consisted levels of DNA methylation, we found, in line with the previous of the strongest changes (supplementary fig. S3C, observations in human, an even clearer trend of increased Supplementary Material online). These results suggest that DNA methylation in cancer (fig. 3A). The PCA hinted at some there are important differences in the breadth and distribu- differences between the phenotypes in that a slight separa- tion of hyper- and hypomethylation alterations in cancer, tion was observed between the groups when looking at the whereas aging-associated alterations appear to be located at top two dimensions (fig. 3B). The differential methylation similar genomic locations regardless of the direction of analysis (FDR < 0.05, average change in methylation >0.2) change, these locations being comparable to those targeted uncovered extensive alterations, with 181,027 dmCpGs by cancer-associated hypomethylation. detected in cancer and 22,564 in aging (lists of dmCpGs are expand on the possible functional consequences of the available as extended data sets). When all of the nine non- DNA methylation alterations observed, we mapped the tumoral samples (nontumor, young, and old groups) were dmCpGs to genes and performed pathway enrichment anal- used as the control group in the cancer differential methyla- yses. First, as shown in figure 2F, the largest proportion of tion comparison, results and numbers of dmCpGs were sim- dmCpGs mapping to genes was found for cancer hyperme- ilar (supplementary fig. S5A, Supplementary Material online), thylation. However, of the CpGs that mapped to genes, can- suggesting that the cancer-associated alterations observed are cer hypo-dmCpGs and aging dmCpGs, although less readily detected irrespective of the samples used. Regarding numerous, were distributed across similar numbers of genes the direction of the alterations, hypermethylation was the as cancer hyper-dmCpGs (fig. 2G, v2 P < 0.001), indicating predominant phenomenon in both processes (accounting that the former were more spread-out across different genes. for 68% and 62% of changes, respectively, fig. 3C and D). We performed pathway enrichment analyses against several The genomic-location distributions of the dmCpGs were databases which included Gene Ontology, KEGG and analogous to those observed in the human system (fig. 3E): Reactome pathways, and the CGP database from MSigDB, cancer hypermethylation occurred at CpG island-associated which describes empirical pathways mined from biomedical sites whereas cancer hypomethylation and aging hyper- and literature. Only cancer hyper-dmCpGs revealed significant hypomethylation changes were much more enriched at open enrichment in biological functions, and these were related sea locations (Fisher’s P < 0.001, ORs ¼ 24.7, 6.7, and 10.3). to Polycomb-target genes, G-protein coupled receptor Similarly, with respect to gene locations, these latter altera- (GPCR) pathways—with neuropeptide receptors in particu- tions were once more enriched at intergenic and intronic lar—and cellular development pathways, especially neuronal sequences (fig. 3F, Fisher’s P < 0.001, intergenic ORs ¼ 7.5, (supplementary fig. S4, Supplementary Material online, full 2.4, and 3.0, intronic ORs ¼ 1.6, 2.3, and 2.1, respectively). results are available as extended data sets). Thus, cancer- Again, cancer hypermethylation was directly associated with associated DNA methylation deregulation, particularly hyper- higher CpG-density (supplementary fig. S5B, Supplementary methylation, appears to have a more coherent functional Material online) and accounted for the largest alterations impact than aging-associated changes. (supplementary fig. S5C, Supplementary Material online). In In order to investigate common axes of DNA methylation general, the patterns of disease-associated DNA methylation deregulation in both cancer and aging, we intersected the changes, and especially the contrast between cancer and ag- dmCpGs found, observing common CpG sets across all pro- ing alterations, followed the same trends as those seen for cesses (fig. 2H and supplementary table 4, Supplementary human tissue. Material online). These intersections were all significantly examine if the parallelisms observed extended to the func- overenriched (SuperExactTest all P < 0.001), with common tional pathways involved, we mapped the dmCpGs to genes cancer- and aging hypo-dmCpGs displaying the highest fold and carried out pathway enrichment analyses. Once more, enrichment, whereas common cancer hyper- and aging hypo- within the dmCpGs mapping to genes, despite cancer hyper- dmCpGs showed the lowest. These findings indicate that, dmCpGs being by far the most numerous, they were concen- regardless of their direction of change, DNA methylation trated across fewer genes than expected when compared alterations are readily found at disease-common loci. We with hypo-dmCpGs and aging changes (fig. 3G, v2 3420 Aging and Cancer Epigenetic Signatures across Human and Mouse . doi:10.1093/molbev/msab112 MBE A B C D E F G H FIG. 3. Cancer and aging DNA methylation alterations in mouse. (A) Violin plots of global levels of DNA methylation of all CpGs across the profiled samples (N, nontumor; T, tumor; Y, young; O, old). (B) PCA plots showing the distribution of the samples across the first two principal components, with the percentage of variance each explains. (C) Heatmaps showing the methylation values of the top 1,000 dmCpGs found in the aging and cancer comparisons. (D) Barplots indicating the relative numbers of hyper- and hypo-dmCpGs (differentially methylated CpGs) in aging and cancer. (E and F) Barplots indicating the distribution of the dmCpGs across CpG island and gene-region locations, including the distribution of the universe of CpGs profiled by the RRBS (“all”). (G) Lollipop plots comparing the number of dmCpGs mapped to genes (left axis, black dots) with the number of different genes that the dmCpGs mapped to (right axis, orange dots). (H) UpSet plot describing the dmCpG sets and their intersections, indicating the size of the sets and the fold-enrichment (FE) of the intersections based on their expected overlaps. P < 0.001). In this case, the pathway enrichment analyses available as extended data sets). Cancer hyper-dmCpGs, as revealed significant results for all sets of dmCpGs (supplemen- with human, were enriched in Polycomb-target genes, GPCRs tary fig. S6, Supplementary Material online, full results are (Wnt-activated in particular), and cellular development 3421 Perez et al. . doi:10.1093/molbev/msab112 MBE pathways, including neuronal ones. On the other hand, hypo- different to those found for cancer. We observed enrichments dmCpGs were enriched in generic cancer pathways, cell- at locations marked with gene-body H3K36me3, heterochro- signaling pathways (including GPCRs), and inflammation matin H3K9me3, and enhancer/promoter H3K4me1. In this terms. Aging hypermethylation was enriched in cell- case, even though we observed an increase in H3K4m1 asso- interaction terms, and hypomethylation in cell-signaling ciation at aging hypomethylated loci when compared with pathways. These results indicate that, in contrast to human aging hypermethylation—a finding reported for human in processes, aging-associated DNA methylation alterations in several tissues (Fernandez et al. 2015; Perez et al. 2018)— mouse may have more directed functional consequences at the global patterns of enrichment found for aging dmCpGs the gene level. differed moderately from those described in the literature, a We integrated the cancer- and aging-associated dmCpGs result which could have been influenced by the fact that the and found significant overenrichment for all the common “classical” chromatin signatures have been mostly derived by sets (fig. 3H and supplementary table 4, Supplementary using array technology, which targets similar but not identical Material online, SuperExactTest all P < 0.001). As occurred genomic contexts (supplementary fig. S7, Supplementary with human tissue, whereas all the intersections displayed Material online). In further support of this idea, a recent study overenrichment, common cancer-, and aging hypo- comparing aging-associated DNA methylation changes in dmCpGs displayed the highest fold enrichment and cancer liver between human and mouse—using array and RRBS hyper- along with aging hypo-dmCpGs the lowest. Next, we data for each species respectively—reported a high enrich- performed pathway enrichment analyses of the intersections. ment in H3K36me3-associated loci specifically for mouse ag- This time, we observed no significant enrichments for those ing changes as compared with human (Wang et al. 2017). containing cancer hyper-dmCpGs, but modest enrichments Because the histone modification results exposed epige- for inflammation terms did appear, especially for cancer nomic similarities between human and mouse for the meth- hypo- and aging hyper-dmCpGs (full results are available as ylation changes associated with cancer and aging, we extended data sets). Taken together, these results suggest extended our analyses to discover and study chromatin that, in the case of mouse and, in contrast to the human states, which are defined by spatial combinations of the his- results, the Polycomb-target, developmental gene hyperme- tone marks and reflect functional genomic states (Ernst and thylation observed in cancer is not as evident in aging. Kellis 2010). We trained 18-state models using the six core histone modifications for our selected panel of data sets by The Interspecies Genomic Context of Cancer and using multivariate Hidden Markov Models (see Materials and Aging DNA Methylation Alterations Methods). The nature of the human states defined was sim- In light of the parallelisms observed, we sought to extend ilar to that previously described (Roadmap Epigenomics the analyses to additional levels of epigenomic regulation Consortium et al. 2015) in terms of their histone mark com- by integrating the methylomic data with the previously position and co-occurrence with other genomic elements mentioned histone modification data—involving the (supplementary fig. S8A, Supplementary Material online). histone modifications H3K4me1, H3K4me3, H3K27ac, The states found for mouse were, in turn, notably similar to H3K36me3, H3K27me3, and H3K9me3—across a panel human (supplementary fig. S8B, Supplementary Material on- of common healthy tissues in human and mouse (see line), with the exception of some specific ones, such as the Materials and Methods). We performed overenrichment one assigned as “EnhG2” (genic enhancer 2). These observa- analyses (FDR < 0.05) to test the association of the cancer- tions suggest that multilevel epigenomic regulation is func- and aging-related dmCpGs discovered in the two species with tionally parallel between the two species. the different histone modifications, in order to define chro- We evaluated the overenrichment of the dmCpGs in the matin signatures linked to the DNA methylation changes. different chromatin states (fig. 4B and supplementary table 6, Using this procedure, we were able to find robust and well- Supplementary Material online). Cancer hypermethylation defined similarities between the two species (fig. 4A and sup- was strongly enriched in bivalent transcription start sites plementary table 5, Supplementary Material online). and Polycomb-repressed sites (states 14, 16), and somewhat As regards cancer-associated changes, we observed an asso- enriched in transcription start sites and enhancer sites in ciation of DNA hypermethylation with the repressive general (states 2–4, 7–11) for both species. Additionally, H3K27me3 modification and the more active enhancer/ and as suggested by the stronger enrichment in H3K9me3 promoter-associated H3K4me1. In the case of human, cancer for human as compared with mouse (fig. 4A), the chromatin- hypermethylation was also strongly enriched at H3K9me3 state analysis revealed that cancer hypermethylation also oc- locations, which are linked to heterochromatic and repressive curred, specifically for human, both at heterochromatic states. On the other hand, cancer hypomethylated dmCpGs regions and at zinc-finger (ZNF) genes/repeats (states 12, were mainly enriched in the H3K9me3 mark for both species. 13). Indeed, ZNF genes, which are characterized by the high These cancer-associated signatures have been described presence of H3K9me3 (Blahnik et al. 2011), displayed the across multiple tissues for human (Perez et al. 2018) whereas highest overenrichment of all the chromatin states in human the literature for mouse is scant. cancer hypermethylation. Interestingly, this family shows a en looking at aging-related signatures, we again found great evolutionary expansion in human as compared with comparable patterns between human and mouse. other species, including mouse, and its members are posited Moreover, for both species, these patterns were markedly to play roles in novel transcriptional repression activity 3422 Aging and Cancer Epigenetic Signatures across Human and Mouse . doi:10.1093/molbev/msab112 MBE A B C D FIG. 4. Chromatin landscape of aging and cancer DNA methylation alterations in human and mouse. (A and B) Heatmaps showing the significant overenrichments in log2(odds ratio) of hyper- and hypo-dmCpGs in the cancer and aging comparisons, for human and mouse in ChIP-seq peak locations of different histone marks across a panel of tissues or in chromatin states built across the same panel of tissues. (C) Bubble plots showing the significant overenrichments in log2(odds ratio) of hyper- and hypo-dmCpGs in the cancer and aging comparisons in the locations of repetitive DNA classes across the genome for human and mouse. (D) Barplots indicating the relative numbers of transcription factors, classified by their domains, whose motifs were found to be significantly enriched (HOMER analysis, FDR<0.05) at the genomic locations of the different dmCpG sets, compared with the numbers of all of the transcription factors tested (“all”). Only the top 50 most enriched, significant factors for each set were selected. (Emerson and Thomas 2009). Thus, the DNA methylation hypomethylation, for both species, most of the affected chro- deregulation observed in human cancer could be related to matin states (13, 15, 17, 18) involved repressive functions, the targeting of human-specific pathways. Regarding cancer such as heterochromatin or Polycomb-associated sites, with 3423 Perez et al. . doi:10.1093/molbev/msab112 MBE some enrichment also being observed in various enhancer dmCpG sites, selecting the top 50 significant (FDR < 0.05) associated states. In the case of aging-associated dmCpGs, motifs with the highest enrichments found for each compar- and much like with the histone signatures, the enrichments ison (see supplementary table 8, Supplementary Material on- were spread out across similar states for hyper- and hypome- line, for extended results). First, to obtain a general overview thylation, and involved gene-body transcription states (5, 6), of the types of TFs affected by the DNA methylation changes, enhancer states (7–11), and repressive states (15–18). we classified our results into TF domain-families (fig. 4D). Especially in the case of human, a particular increase in en- With this approach, we found both similarities and differ- richment in enhancer states was observed for aging hypome- ences between the processes as well as between species. thylation. The human and mouse profiles were globally Cancer hypermethylation dmCpGs were associated with similar, with some differences, such as in EnhG2 (state 8) Homeobox TFs for both species (Fisher’s P < 0.01, probably being explained by the pre-existing differences in ORs ¼ 2.7 and 2.7 for human and mouse, respectively), but the chromatin models built (supplementary fig. S8, also associated with NRs in human (Fisher’s P < 0.05, Supplementary Material online). Collectively, our results OR ¼ 2.8) and bHLH-domain TFs (Fisher’s P < 0.001, show that the chromatin context of DNA methylation OR ¼ 7.8) in mouse. Regarding cancer hypomethylation, it changes in cancer and aging is considerably different, and was particularly enriched at bHLH- and HMG-domain TF that the specific features of each process are robustly recapit- motifs in human (Fisher’s P < 0.001, ORs ¼ 11.4 and 8.8, ulated in human and mouse. respectively). On the other hand, the patterns of TF sites We also looked at the association between DNA methyl- targeted by aging hyper- and hypomethylation were similar ation alterations and repetitive DNA elements (fig. 4C and in both species. Aging hypermethylation particularly affected supplementary table 7, Supplementary Material online; see ETS-domain TF motifs (Fisher’s P < 0.001, ORs ¼ 48.5 and supplementary fig. S9, Supplementary Material online, for the 26.9 for human and mouse, respectively) whereas hypome- analysis expanded to repetitive DNA classes). First, regarding thylation involved bHLH- and HMG-domain TF sites (Fisher’s cancer, repetitive loci were mainly associated with hypome- P < 0.001, ORs ¼ 6.9, 12.0, 7.8, and 16.6 for human bHLH, thylation across the same large classes of repeats (SINE, LINE, HMG and mouse bHLH, HMG, respectively). LTR, DNA), as has been well-described previously (Ross et al. Subsequently, we focused on the specific TFs involved in 2010), with the main species difference being that satellite the previous observations, by looking at the main sets of repeats showed no changes in DNA methylation for mouse shared or unique TFs affected in each comparison (supple- (fig. 4C). In the case of SINE repeats, both the primate-specific mentary fig. S10, Supplementary Material online). As sug- Alu family in human and the related B elements (B1, B2, B4) in gested by the previous results, the patterns of species- mouse were targeted by DNA hypomethylation (supplemen- common alterations at TF motifs were especially noticeable tary fig. S9, Supplementary Material online). Interestingly, we for aging-related changes. Aging-associated hypermethylation also observed in human the hypermethylation of ribosome- of ETS-domain TFs, a family strongly linked to oncogenic associated elements (rRNA, srpRNA), a finding which has processes (Sizemore et al. 2017), was the main common epi- been reported for some types of cancers (Srivastava et al. genetic phenomenon between mouse and human. Several 2016), although literature on the subject is scarce. Secondly, Sox TFs (HMG domains) were commonly targeted by in the case of aging, both hyper- and hypomethylation were aging hypomethylation in both species, such as Sox2, whose associated with repetitive DNA, with the patterns observed expression has been shown to be deregulated with aging being similar to those of cancer hypomethylation. Although in both human and mouse brain (Carrasco-Garcia 2018). RRBS is mostly limited to examining single-copy sequences, Interestingly, Sox TFs also appeared affected by cancer hypo- the bidirectionality of the changes found here for aging may methylation in human, suggesting possible links between the help explain observations in the literature that fail to describe regulation of aging and of cancer, as has been shown for Sox4 any noticeable trend of loss of methylation during aging in in other tumors (Foronda et al. 2014). In addition, the dereg- the genome-wide levels of brain methylation both in mouse ulation of methylation patterns at bHLH-domain TF sites (Hadad et al. 2019) and human (Lister et al. 2013; McKinney appeared to be a general trend in both cancer and aging et al. 2019). No associations whatsoever were found with processes, appearing at loci related to specific neuronal de- simple or low-complexity repeats, which were the main clas- velopment and differentiation, including TFs such as ses which we had previously linked to low levels of basal NeuroG2, NeuroD1, Olig2, or Atoh1 (Dennis et al. 2019) methylation in nontumoral tissue (fig. 1J). Taken together, and also involving Homeobox TFs such as DLX1/2. these observations suggest that repetitive DNA elements To sum up, the analysis of the links between DNA meth- are subject to similar DNA methylation alterations in cancer ylation alterations and TF-binding sites revealed that clear and aging for both species, with the exception of satellite functional parallelisms between human and mouse exist in repeats, which appear to be more affected by these processes the aging scenario. This is in contrast to our earlier gene in human as compared with mouse. pathway enrichment analyses, which showed that most of We next sought to study the possible effects of the DNA the affected genetic pathways common in human and mouse methylation changes found at TF-binding sites, because TFs involved cancer hypermethylation (of Polycomb-targets and are downstream genomic regulators directly in contact with developmental genes). This discrepancy could be explained the genomic structure. To this end, we performed a HOMER by aging-associated methylation changes being more func- analysis to find enrichment in TF motifs associated with tionally important when occurring at genomic sites not 3424 Aging and Cancer Epigenetic Signatures across Human and Mouse . doi:10.1093/molbev/msab112 MBE directly associated with gene locations (such as many TF- observation, we directly assessed the transitions between the binding sites). Moreover, as shown in earlier stages of this genomic contexts of the sdCpGs in the genome of each spe- study (figs. 2F and 3F), aging changes concentrate more at cies (fig. 5E), finding that, particularly in the case of hyper- nongenic sites than cancer changes. Thus, the identification of sdCpGs, there was a considerable difference in the genomic functional targets implicated in DNA methylation alterations association to CpG island status between the two species. In in aging may be more effective when focusing on TF-binding this case, CpGs which in human-occupied CpG islands and in site alterations. mouse-occupied less-dense regions (shores, open sea) had higher methylation levels in mouse (i.e., were classified as Interspecies Integration of Cancer and Aging DNA hyper-sdCpGs). This is in line with our previous observations Methylation Alterations (fig. 1F–H) of denser genomic loci being associated with lower After establishing the parallel features of DNA methylation methylation values in both species. This finding suggests alterations in human and mouse, we integrated our CpG- that interspecies differences in DNA methylation levels specific data, to be able to directly compare methylation may be partly explained by local changes in the functional values between the two species at orthologous locations, by genomic context of their respective genomes, such as CpG lifting the mouse coordinates to the human genome (see island membership (Hernando-Herraez, Garcia-Perez, et al. Materials and Methods). We recovered interspecies measure- 2015). ments for a sizeable set of 59,100 CpG sites, 35% of which had xt, we focused on the fraction of previously described been classified as dmCpGs at any comparison (see supple- species-specific cancer- or aging-associated dmCpGs for mentary table 9, Supplementary Material online, for the lift- which we had interspecies measurements (see supplementary over statistics; the full list of orthologous sites is available as an table 9, Supplementary Material online, and extended data extended data set). These orthologous sites were overen- sets). This allowed us to directly compare cancer- and aging riched in CpG island locations when compared with the orig- changes at specific loci between human and mouse. We inal backgrounds (Fisher’s P < 0.001, ORs ¼ 7.1 and 6.1 for found common and statistically significant intersections be- human and mouse, respectively), and largely maintained their tween the two species for all sets (fig. 5F and supplementary relationship to CpG islands in both species, suggesting that table 10, Supplementary Material online; SuperExactTest all they covered similar functional regions in both genomes P < 0.001), with the biggest intersection being found for (fig. 5A). Correlation analysis of the DNA methylation values cancer dmCpGs, although the observed size and fold enrich- of the interspecies lifted CpGs revealed that, excluding cancer, ment of the intersections was probably influenced by the the samples clustered within their species (fig. 5B). Tumors, differing sizes of the various sets. Differentiating by direction on the other hand, again showed widespread reconfiguration, of change revealed that the biggest intersections were dom- clustering independently and, interestingly, mouse tumor inated by hypermethylation changes in both species (supple- CpGs were slightly more correlated to human tumors than mentary fig. S12, Supplementary Material online), especially to the rest of the mouse samples. Examining only nontumoral when involving cancer sets. Aging, on the other hand, was samples, methylation was found to be highly correlated be- limited to much smaller intersections with directionality not tween the two species (fig. 5C, left panel; Pearson coef. 0.84, playing such an apparent role. Nonetheless, the observed P < 0.001). Moreover, when we retained only CpG sites which intersections suggest that, added to the previously described were lowly variable within each species (SD < 0.1 methylation similarities in genomic contexts (fig. 4), when we focus on value) most of the species-discordant CpG sites disappeared specific orthologous loci, common sites of DNA methylation (fig. 5C, right panel; Pearson coef. 0.92, P < 0.001). Indeed, the alteration across the two diseases and the two species are subsets of high-variability sites for human and mouse (CpGs readily detected. with SD 0.1 methylation value) showed much higher over- After the characterization of specific similarities between lap than expected (fig. 5D; Fisher’s P < 0.001, OR ¼ 16.7), human and mouse, we turned to examine the influence of suggesting that locations of methylation variability are con- sequence conservation in the explanation of the epigenetic served between the two species. In order to strictly define patterns found. We integrated our data with UCSC conser- species-discordant CpGs (sdCpGs), we performed a differen- vation scores for multiple alignments across vertebrate spe- tial methylation analysis (FDR < 0.05, average change cies (see Materials and Methods). First, regarding disease- in methylation >0.2) at the 59,100 loci with interspecies associated alterations in general (dmCpGs), we observed measurements (the list of sites and sdCpGs is available as that, for both species, cancer hypermethylation occurred at an extended data set). We identified 1,845 sdCpGs (supple- more conserved regions than did either cancer hypomethy- mentary fig. S11A, Supplementary Material online), evenly lation or aging changes (fig. 5G). This finding can support the split between those presenting higher and those lower meth- notion that cancer hypermethylation may impact conserved ylation in mouse as compared with human (hyper- and hypo- axes of gene regulation across species, something which we sdCpGs, respectively, supplementary fig. S11B, Supplementary had observed through the pathway enrichment analyses. Material online). Interestingly, when annotating the sdCpGs However, these trends may also be partly explained by the either to human or mouse genomic contexts, we observed genomic distribution of the alterations, because hypomethy- considerable differences in the distribution of the sites, espe- lation and aging alterations occur more often at intergenic cially according to their CpG island status (supplementary fig. and intronic regions (see figs. 2F and 3F), which we found S11C and D, Supplementary Material online). Following this (using all the profiled CpGs) to be less conserved than other 3425 Perez et al. . doi:10.1093/molbev/msab112 MBE A B C D E F G H FIG. 5. Interspecies DNA methylation alterations in human and mouse. (A) Sankey diagram describing the distribution of changes in the CpG island status of the 59,100 interspecies orthologous CpGs when considering either the human or the mouse genomic annotation. On either side, the original backgrounds of all the profiled CpG sites are shown. (B) Heatmap showing the pairwise Pearson correlations of the DNA methylation levels of the interspecies CpGs across all samples. (C) Scatter plots showing the correlation between DNA methylation levels at all the profiled interspecies sites (left plot) or excluding intraspecies highly variable CpG sites with SD 0.1 methylation value (right plot) (tumors excluded). (D) Venn diagram showing the intersection of interspecies lifted CpG sites classified as highly variable within each species (SD 0.1 methylation value) for both mouse and human. (E) Sankey diagrams showing the distribution of hyper- and hypo-sdCpGs (species-discordant CpGs) across CpG island locations, taking into account the differences in the genomic context of the human and mouse genome. (F) UpSet plot describing the sets of cancer and aging dmCpGs with interspecies measurements, showing their sizes and intersections between human and mouse. (G) Violin plots showing the distribution of phastCons conservation scores for the detected aging- and cancer dmCpGs in human and mouse, compared with the profiled RRBS background (red dots indicate average values). (H) Violin plots showing the distribution of phastCons scores for the detected interspecies sdCpGs, compared with the 59,100 lifted background, either in human or mouse-centered conservation score (red dots indicate average values). 3426 Aging and Cancer Epigenetic Signatures across Human and Mouse . doi:10.1093/molbev/msab112 MBE genetic regions such as exons or promoters, that is, those upstream and downstream of many alternative transcripts more impacted by cancer hypermethylation in both species (fig. 6A, upper plots), whereas DNA methylation alterations (supplementary fig. S13A and B, Supplementary Material on- in the same direction, albeit smaller, also appeared with aging line). Second, within our 59,100 interspecies loci, we found in both species (fig. 6A, lower plots). The specific location of that sdCpGs were enriched at conserved regions (fig. 5H) the DNA methylation changes found hinted at a nonrandom more frequently than the rest of the interspecies sites. This and perhaps functional role for the alterations. Thus, we an- observation could at first sight appear to challenge the exist- alyzed AGAP3 expression in the TCGA LGG-GBM and found ing literature, which describes the positive correlation be- a strong decrease in expression in tumors (fig. 6B, Wilcoxon tween sequence conservation and epigenetic conservation rank-sum P < 0.001). Moreover, expression levels were (Zhou et al. 2017), but, again, it may partly be explained strongly associated with overall survival in the cohort by the sdCpGs being enriched at exon locations (see supple- (fig. 6C, log-rank test P < 0.001), with the more lethal GBM mentary fig. S11D, Supplementary Material online), which are subtype having the lowest levels of expression and survival more conserved (see supplementary fig. S13B, Supplementary (supplementary fig. S15A and B, Supplementary Material on- Material online). To address these issues, we applied stratified line). Interestingly, as can be seen in figure 6A, the AGAP3 sampling to compare the CpG sets of interest against ran- internal hypermethylation events colocated with the ending domly sampled background sets of CpGs which had match- and start of several isoforms. We therefore explored the levels ing distributions of gene locations. With this strategy, we of the main AGAP3 isoforms (supplementary fig. S15C, found that most of the trends regarding the difference in Supplementary Material online) and their association with conservation scores observed could indeed be partly survival (supplementary fig. S15D, Supplementary Material explained by the genomic context of the CpGs involved, online), discovering that a specific isoform (uc003wjf/ both for cancer- and aging dmCpGs (supplementary fig. ENST00000473312) was mostly explanatory of the effect (sup- S14A, Supplementary Material online) and for human- and plementary fig. S15E, Supplementary Material online, Cox mouse sdCpGs (supplementary fig. S14B, Supplementary Models log-rank test P < 0.001). Moreover, it corresponded Material online). These results suggest that both disease- to an isoform truncated close to the internal region targeted and species-specific epigenetic traits are influenced by by the DNA methylation alterations (highlighted in fig. 6A). the genomic context surrounding the specific CpG sites Indeed, studies have described isoform variants of AGAP3 involved. with biological functions (Oku and Huganir 2013) some of which may have roles in oncogenic processes through alter- Interspecies Hotspots of Cancer- and Aging Epigenetic ations in their mRNA expression (Shimizu et al. 2019). These Deregulation observations exemplify an interspecies DNA methylation de- Finally, we used the interspecies CpGs to look for common regulation process which, as shown for human, could have a hotspots of cancer- and aging epigenetic deregulation be- role in AGAP3 deregulation in glioma, possibly through an tween human and mouse which could have downstream effect on isoform switching, with aging-associated methyla- functional implications. We focused on the 22 CpGs which tion changes preluding the cancer-associated alterations. had been identified as dmCpGs in both species and both addition to AGAP3, five other genes were associated with processes (described in fig. 5F). Of these, nine CpG sites pre- concordant interspecies dmCpGs (fig. 6D): GPRIN1, AJM1, sented concordant changes in both species (i.e., being deregu- LHX2, CCDC177, and TPGS1. Notably, four of the genes iden- lated in the same direction, in cancer and aging, for human tified, much like with AGAP3, were clearly downregulated in and mouse) and were associated with a total of six genes (see tumoral samples as compared with healthy tissue (fig. 6E, supplementary table 12, Supplementary Material online). In Wilcoxon rank-sum tests), with the GBM subtype often dis- order to study the possible functional consequences of the playing the strongest downregulation. We also investigated DNA methylation alterations, we obtained TCGA expression their relationship with survival and found a strong association data for the LGG–GBM glioma cohorts and analyzed the between gene expression and overall survival for GPRIN1, with expression of these candidate genes and their association a more modest effect for AJM1 and, interestingly, a trend for with survival across glioma patients (see Materials and an association between CCDC177 expression and survival that Methods). Notably, we found evidence of the strong deregu- was specific to the GBM subtype (supplementary fig. S16, lation of five out of the six genes in glioma samples as com- Supplementary Material online). LHX2 is involved in neural pared with controls (see below), indicating that our robust development (Chou and Tole 2019), whereas GPRIN1 and approach of looking at orthologous and concordant aging- TPGS1 play a role in neural cytoskeleton dynamics (Regnard cancer DNA methylation changes leads to the discovery of et al. 2003; Nordman and Kabbani 2012) and AJM1 is related functionally relevant targets. to apical junctions (Köppen et al. 2001). As such, the deregu- We first identified the AGAP3 gene as being a target of lated genes appear to be predominantly associated with cy- both cancer-associated and aging-associated deregulation in toskeletal processes. LHX2 is known to be oncogenic in several both human and mouse. AGAP3 is a component of the neural types of cancer (Song et al. 2019) although we have found it NMDA receptor complex and AGAP proteins regulate recep- to be downregulated in the case of glioma, as has been tor trafficking in neurons (Oku and Huganir 2013). In cancer, reported previously (Cheng et al. 2019). Nonetheless, the this gene showed strong hypermethylation of internal CpG specific role of LHX2 and of the rest of the genes in glioma sites which, in both species, colocated with CpG islands 3427 Perez et al. . doi:10.1093/molbev/msab112 MBE A B D C E FIG. 6. AGAP3 is an interspecies cancer- and aging-associated target of DNA methylation alterations. (A) Plots describing the average methylation values of tumoral, nontumoral, young, and old samples in human and mouse across the profiled CpG sites mapping to AGAP3. Sites marked in yellow represent significant dmCpGs in the aging and cancer comparisons. Below, the transcripts and CpG islands associated with the genomic region described are shown. (B) Boxplots showing the expression values of the AGAP3 gene in control and tumor samples from the TCGA glioma cohort (GBM-LGG). Expression is measured in VST (variance stabilizing transformation) normalized units. High and low-expression samples (above and below the median) are in yellow and green. (C) Survival curves showing the association of high and low AGAP3 expression groups with overall survival across glioma patients. (D) Boxplots describing the methylation values of the interspecies dmCpGs that presented concordant changes in cancer and aging across human and mouse and which were associated with specific genes. (E) Boxplots showing the expression values (VST normalized units) of the genes associated with the interspecies concordant dmCpGs in control and tumoral samples from the TCGA glioma cohort (GBM-LGG). Glioblastoma (GBM) and lower grade glioma (LGG) samples are shown separately. High- and low-expression samples (above and below the median) are color-coded. Wilcoxon rank-sum tests: *P<0.05, **P<0.001, ***P<0.0001, “ns,” P>0.05. 3428 Aging and Cancer Epigenetic Signatures across Human and Mouse . doi:10.1093/molbev/msab112 MBE described here, or in cancer as a whole, remains largely across tissues (Michalak et al. 2019), it will be of value to unexplored. explore the interspecies conservation of epigenetic signatures in other tissues (Maegawa et al. 2017; Wang et al. 2017), and also other species such as pig or rhesus. Third, we used RRBS Conclusions to profile analogous genomic locations in human and mouse, In this work, we have performed a systematic analysis of the a technology which focuses on CpG-dense locations and is DNA methylation dynamics associated with cancer and aging more limited when studying intergenic regions. That said, the in brain tissue in human and mouse. We employed RRBS as a use of RRBS allowed us to define a large subset of interspecies platform which allowed us to profile comparable genomic measurements (59,100 CpG sites) with adequate coverage locations in both species. With this strategy, we first con- across all the samples. firmed, in our study system, the parallelisms in baseline From an evolutionary perspective, systematic multispecies DNA methylation patterns between human and mouse. comparisons are of great value in the understanding of the Subsequently, we explored the alterations associated with etiology of disease. The uncovering of species-common path- cancer and aging in both species. Our analyses revealed ways provides robust molecular explanations for the mecha- that there are considerable differences in the genomic and nisms at play, whereas interspecies differences may help epigenomic features of DNA methylation alterations in can- explain the phenotypic divergences observed across the spe- cer and aging, and that the particular characteristics of the cies. For instance, our observation of the human-specific epi- alterations found for each process are globally conserved in genetic deregulation of zinc-finger genes in cancer may be human and in mouse (see supplementary fig. S17, related to the importance of this family in primates, which Supplementary Material online, for a graphical outline of have experienced considerable lineage-specific expansion as the main findings). We found specific parallelisms at the level compared with rodents (Huntley et al. 2006; Emerson and of their genomic locations and the magnitude of the changes, Thomas 2009). The emphasis on species-specific pathways, the genetic pathways involved, the chromatin context of the whereas relevant to evolutionary biology is, conversely, of less alterations, the families of repetitive DNA elements affected, value in clinical biology, which seeks the translation of mech- and the alteration of TF regulation sites, among others. anisms to human. In this respect, systematic, genome-wide, Interestingly, we observed that aging-associated changes in multispecies integrative studies such as the one presented general seem to impact more specific functional gene path- here have the potential to answer questions relating to ways in mouse than in human. Nonetheless, we found that both ends. the classical hypermethylation of developmental gene path- Taking a more clinical viewpoint, we can foresee that our ways is more shared between cancer and aging in human results may spark interest in different areas. The mouse is one than in mouse. Intriguingly, epigenetic alteration of satellite of the main preclinical models for human disease and, as such, repeats appeared to be a feature of both aging and cancer in it is essential to have markers for the prediction of treatment human, but not in mouse. Significantly, we found very clear success in human (Day et al. 2015). In this vein, in this work, interspecies similarities in the aging-associated alterations oc- we found that species-common epigenetic changes across curring at the TF-binding sites of specific families, suggesting diseases were readily associated with genes showing expres- that the functional impact of aging methylation changes sion alterations in glioma in human. Thus, focusing on con- could indirectly occur through TF deregulation rather than served epigenetic alterations such as those described here through the direct targeting of specific genes. Finally, we de- may help narrow the search for functionally relevant targets rived a large set of orthologous CpG sites with interspecies that are altered in disease. In this sense, current advances in measurements, which allowed us to observe that sites of the development of targeted epigenetic editing tools (Liu epigenetic variability are conserved between the species, et al. 2016), combined with the knowledge of epigenetic con- that interspecies differences between human and mouse servation in mouse and human, may serve as a powerful tool can be partly explained by changes in the local genomic con- to explore the overall impact of both epigenetic and pathway text associated with the CpG sites and that there are com- alterations in multiple biological scenarios. mon interspecies loci which are altered during both aging and On the other hand, the set of epigenetically conserved cancer. DNA methylation sites could pave the way for the design The results presented in this manuscript cover several and development of novel custom high-throughput plat- aspects of the epigenetic dynamics in human and mouse, forms, either in the context of cost-effective microarray tech- but we should also draw attention to the limitations of our nology or through the use of more sophisticated targeted study. First, we recognize that the sample size used in our bisulfite sequencing approaches. This may be of particular study design is limited (n ¼ 3 per group), which is principally interest within the framework of epigenetic profiling during because of the substantial number of groups being analyzed drug screening in preclinical models as alterations of such (four phenotypes in two species). However, the interspecies conserved disease-associated loci could help predict future corroboration of the results does indicate that the observed treatment success in human. For example, glioma is a cancer patterns are robust. Second, we have focused on a single type with a high prevalence of isocitrate dehydrogenase tissue (brain). Although this allowed us to perform a com- mutations which lead to important epigenetic alterations parative study in greater depth, and it is known that epige- (Han et al. 2020). The main chemotherapy treatment in gli- netic alterations that arise in cancer or in aging are similar oma involves temozolomide (Hirst et al. 2013), and its 3429 Perez et al. . doi:10.1093/molbev/msab112 MBE effectiveness is also associated with various DNA methylation and female individuals, sex chromosomes were not included features beyond the well-known MGMT promoter methyla- in the subsequent analyses for any species (see below) in tion (Cheng et al. 2019). Thus, integrating epigenetic conser- order to avoid sex bias. vation knowledge could aid in the identification of drug- Human samples and data from patients included in this sensitive pathways in animal models, which in turn would study were provided by the following institutions: the improve the current standard-of-care treatments as well as Biobank HUB-ICO-IDIBELL (PT17/0015/0024, integrated in also shed light on the mechanistic insights of novel com- the Spanish Biobank Network), the Biobank of Galicia Sur pounds or those rescued from current drug repositioning Health Research Institute (IISGS), and the Aragon Health approaches. Sciences Institute in the framework of the Biobank of All in all, this work could constitute a baseline for other Aragon. All human and mouse samples were processed fol- future studies focusing on the systematic evaluation of lowing standard operation procedures with the appropriate disease-associated epigenetic patterns across multiple tissues approval of the corresponding Ethical and Scientific Review and species from a multiomic perspective in order to inform Boards: Bellvitge University Hospital Ethics Committee (code the search for inter- and intraspecies patterns of epigenetic 07/19, April 11, 2019) for the human aging samples; Biobank regulation. IISGS Ethics Committee (code 2019/238, April 28, 2019) and Aragon Clinical Research Ethics Committee (code 14/2019, Materials and Methods July 24, 2019) for the human cancer samples; Health Institute All statistical analyses were carried out using R statistical Carlos III Ethics Committee (code CBA PA 13_2013, February software (v3.6.2) unless stated otherwise. Software was 26, 2013) for the mouse aging samples; Autonomous managed through Bioconda (Grüning et al. 2018). Graphs University of Barcelona Ethics Commission (code CEEAH- were produced using R base functions and the ggplot2 R 3665, February 5, 2018) for the mouse cancer samples. package (v3.2.1) (Wickham 2016). For the illustration and testing of multiple-set intersections, the UpSetR (v1.4.0) and Reduced Representation Bisulfite Sequencing SuperExactTest (v1.0.7) R packages were used (Wang et al. For each sample, 100 ng of genomic DNA were used for li- 2015; Conway et al. 2017). For the illustration of gene-level brary preparation with the Premium Reduced Representation DNA methylation profiles, the ggbio R package (v1.34.0) was Bisulfite Sequencing Kit (Diagenode). Subsequently, samples used (Yin et al. 2012). were pooled in groups of 6 or 8. PCR clean-up after the final library amplification was performed using Agencourt Sample Description and DNA Extraction AMPure XP (Beckman Coulter). The quality of the pools Regarding the aging cohorts, we collected brain cortex tissue was assessed by Qubit fluorometry (Thermo Fisher from three young (24, 33, 35 years; female, male, male) and Scientific) and Fragment Analyzer capillary electrophoresis three old (65, 69, 70 years; female, male, female) human sub- (Agilent). Finally, the library pools were sequenced. Human jects and three young (10 weeks; all female) and three old samples were sequenced on a NovaSeq6000 (Illumina) using (90 weeks; all female) C57BL/6HsdOla/CBA mice. Humans 50-bp paired-end read sequencing (PE50). Mouse samples and mice have different lifespans and in consequence their were sequenced on a HiSeq3000 (Illumina) using 50-bp ages are not equally translatable across their different life single-read sequencing (SR50). stages. The young groups were selected to be at least sexually mature adults, a characteristic which is attained at 10 weeks Reduced Representation Bisulfite Sequencing Data for mice and 20 years for humans (Dutta and Sengupta 2016). Preprocessing Regarding the old groups and considering an average lifespan FASTA quality was assessed with FastQC (v0.11.8; https:// of 80 years and 24 months for humans and mice, respectively www.bioinformatics.babraham.ac.uk/projects/fastqc/ last (Dutta and Sengupta 2016), the 90-week-old mice would accessed April 29, 2021). Reads were quality- and adapter- correspond to 69-year-old humans. Consequently, the age trimmed with Trim galore (v0.6.4; https://www.bioinformat- grouping for both species lies within a similar range. ics.babraham.ac.uk/projects/trim_galore/ last accessed April For the cancer cohorts, we collected glioma and paired 29, 2021) with default parameters under –rrbs mode and – cortex tissue from three human subjects (71, 31, 44 years; 2colour 20 in the case of the human data to account for female, female, male), and glioma and paired cortex tissue NovaSeq bias toward identifying high-quality G bases from from three mouse subjects (12–16 weeks; all female) using an nonsignal basecalls. Next, the reads were aligned to prein- in-house murine model of glioma (see supplementary table 1, dexed and bisulfite converted GRCh38/hg38 and GRCm38/ Supplementary Material online, for additional information on mm10 genomes (primary assemblies) with Bismark (v0.22.2) the samples). The murine glioma model consists in the ortho- using default parameters (Bowtie 2 alignment; see supple- topic allograft of GL261 glioma cells into the brain of C57BL/6 mentary table 1, Supplementary Material online, for addi- mice as previously described by us (Ferrer-Font et al. 2017). tional information on the alignment statistics) (Krueger and DNA was extracted by standard phenol–chloroform Andrews 2011). Finally, methylation counts for cytosines be- procedures. Concentration and quality of the DNA were longing to CpG sites were obtained with Bismark’s methyla- assessed by Qubit fluorometry (Thermo Fisher Scientific) tion extractor under default parameters, including – and Fragment Analyzer capillary electrophoresis (Agilent). no_overlap for paired-end data to avoid scoring overlapping Because the human cohorts were composed of both male methylation calls twice. For each position, methylation values 3430 Aging and Cancer Epigenetic Signatures across Human and Mouse . doi:10.1093/molbev/msab112 MBE were computed as the percentage of methylated cytosines and were accessed through the msigdbr R package (v7.0.1). In with respect to total cytosines on a scale of 0–1. order to retain the most important enrichments, only gene sets with FDR < 0.05, OR >2 and containing between 10 and Differential Methylation Analyses 1,000 genes were represented as statistically significant. Differential methylation analyses were performed on CpG sites, not on individual cytosines, using the following process: Region Enrichment Analyses 1) the counts for cytosines from both strands belonging to Genomic enrichment analyses were performed with the the same CpG site were pooled so as to have one methylation LOLA R package (v1.16.0) (Sheffield and Bock 2016) using measurement per CpG site; 2) CpG sites mapping to sex custom databases. Sets of CpGs were tested for overenrich- chromosomes and alternative haplotypes were filtered out; ment in specific genomic tracks using one-sided Fisher’s exact 3) CpG sites were filtered for low coverage (<10 counts) or tests (FDR < 0.05) with appropriate filtered backgrounds of very high coverage (>99th percentile); 4) CpG sites common the CpGs being used for each case. The genomic regions to all human or mouse samples were retained for the analyses. analyzed included histone marks, chromatin states, and re- Next, the differential methylation analyses were carried out petitive DNA elements. The histone and chromatin state with the methylKit R package (v1.14.2) (Akalin et al. 2012) LOLA databases built for the enrichments are available as using the calculateDiffMeth function with default parameters extended data sets. and multiple testing adjustment by FDR < 0.05. The Histone ChIP-seq BED tracks for H3K4me1, H3K4me3, calculateDiffMeth function uses logistic regression models H3K36me3, H3K27ac, H3K27me3, and H3K9me3 in 15 to define differentially methylated CpGs between groups. healthy human tissues were obtained from LOLA extended Because of the low number of samples (n ¼ 3 per group), a universe (databio.org/regiondb) using NIH Roadmap strict threshold of >0.2 change in methylation value was used Epigenomics region data corresponding to the hg38 version to define significant changes. No other covariates were in- of the human genome (Roadmap Epigenomics Consortium cluded in the models because of the limited information et al. 2015). Histone BED tracks for the aforementioned marks available for other variables. Nonetheless, the inclusion of in nine postnatal mouse tissues mapped to the mm10 version sex or age variables in the human cancer and aging compar- of the mouse genome were obtained from ENCODE (Gorkin isons led to similar results, with 97–99% of the dmCpGs being et al. 2020). the same (data not shown). Chromatin states were built using chromHMM (see below). Consolidated data from the aforementioned six Annotation of Sites to Genomic Locations histone tracks (BED tag.align files) and corresponding ChIP- The profiled CpG sites were annotated to hg38 or mm10 CpG seq input from 15 human tissues were obtained from NIH islands and gene locations using the annotatr R package Roadmap Epigenomics (https://egg2.wustl.edu/roadmap/ (v1.12.1) (Cavalcante and Sartor 2017). Overlapping genetic data/byFileType/alignments/consolidated; last accessed annotations were collapsed with the following priority: April 29, 2021). Data corresponding to the aforementioned Promoter > 5UTR > 3UTR > Exon > Intron > 1– six histone ChIP-seq tracks (BAM files) and the corresponding 5 > Intergenic. Finally, CpG sites were mapped to genes ChIP-seq input from nine mouse tissues were obtained from and their transcripts through the TxDb.Hsapiens.UCSC. ENCODE (https://www.encodeproject.org/; last accessed hg38.knownGene and TxDb.Mmusculus.UCSC.mm10.known April 29, 2021). Gene R packages (v3.1.0). Annotation for the Illumina Repetitive regions identified by RepeatMasker in the hg38 Infinium Human Methylation 450 K Beadchip was accessed or mm10 genomes (Smit 1996) were obtained with the UCSC through the IlluminaHumanMethylation450kanno.ilmn12. Table Browser (Karolchik 2004) and grouped by family or hg19 R package (v0.6.0). class. Elements with an uncertain classification, labeled with To determine the density in surrounding CpGs associated the “?” symbol, were filtered out. with each site, each locus was expanded to a 2,000-bp win- dow and the number of CpGs present in the surrounding Chromatin State Model Learning genomic sequence were counted and divided by the number ChromHMM (v1.18) (Ernst and Kellis 2012) was used to gen- of those possible. erate an 18-state model to predict chromatin states from 15 human and nine mouse tissues. ChIP-seq data sets corre- Gene Set Enrichment Analyses sponding to six histone marks (H3K4me1, H3K4me3, Gene set enrichment analyses were performed with the goseq H3K36me3, H3K27ac, H3K27me3, and H3K9me3), along R package (v1.38.0) (2010 P Genome Biol—Young MD). CpGs with their respective input controls, were obtained from were mapped to genes, and enrichment in biological func- NIH Roadmap Epigenomics and ENCODE for human and tions was carried out with genes that were exclusively hyper- mouse samples respectively (see supplementary table 11, or hypomethylated, using appropriate filtered backgrounds of Supplementary Material online, for the description of the assayed genes. The bias arising from there being different files). Human BED files were converted from hg19 to hg38 number of probes per gene was taken into account in the coordinates via UCSC’s liftOver tool (v377) (Hinrichs 2006). analyses. The GO, KEGG, Reactome, and CGP gene set data- Resulting files were binarized using the binarizeBed function bases from the Molecular Signature Database (MSigDB v7.0) with default parameters using the hg38 chromosome sizes. (Liberzon et al. 2015) were used for the enrichment analyses A model considering 18 states was learned using the 3431 Perez et al. . doi:10.1093/molbev/msab112 MBE LearnModel function with default parameters, as previously broadinstitute.org/runs/stddata__2016_01_28/data/; last described elsewhere (Roadmap Epigenomics Consortium accessed April 29, 2021) and the UCSC Xena (https://xenab- et al. 2015). In the case of mouse data sets (mm10), BAM rowser.net/datapages/; last accessed April 29, 2021) for the files corresponding to replicated samples from different glioma GBMLGG (glioblastoma—lower grade glioma) cohort mouse tissues were merged using SAMtools (v1.7) (Li et al. from the TCGA (The Cancer Genome Atlas) consortium 2009), as previous studies have suggested that chromatin (Brennan et al. 2013; Ceccarelli et al. 2016) (the list of patients state inferences either from single time series or from repli- and genes analyzed is available as an extended data set). Prior cated data are invariably consistent and highly reproducible. to analysis, data were normalized by the VST (variance stabi- Read duplicates were removed using the MarkDuplicates lizing transformation) procedure using the DESeq2 R package function from Picard tools (broadinstitute.github.io/picard/ (v1.26.0) (Anders and Huber 2010). This transformation sta- ). The resulting BAM files were sorted using SAMtools and bilizes the variance–mean relationship in the data and cor- binarized using the binarizeBam function from chromHMM. rects for differences in library size between the samples. For A model considering 18 states was learned using the the isoform analysis, the four most expressed AGAP3 isoforms LearnModel function using the same conditions as the human were selected. Cox models were analyzed with the survival R chromHMM pipeline. For both human and mouse analyses, package (v3.1.11) (Therneau and Grambsch 2000). the resulting chromatin states were reordered and renamed to facilitate later interpretation purposes. In addition, enrich- ment analyses of the different chromatin states in external Supplementary Material genomic data sets (CpG Islands, RefSeq Exons, RefSeq genes, Supplementary data are available at Molecular Biology and RefSeq transcription start sites, RefSeq transcript end sites, Evolution online. and regions within 2,000 bp of a RefSeq TSS) obtained from the UCSC genome browser (hg38 and mm10 genomes) were performed using the OverlapEnrichment function from A nowledgments chromHMM. The authors would like to thank Ronnie Lendrum for man- uscript editing and the members of the Cancer Epigenetics Motif Enrichment Analyses Laboratory for their positive feedback. Finally, they want to Transcription factor motif enrichment analyses were per- particularly acknowledge the patients and biobanks involved formed with Hypergeometric Optimization of Motif in the study: the Biobank HUB-ICO-IDIBELL (PT17/0015/ EnRichment (HOMER) software (v4.11.1) (Heinz et al. 2010) 0024, integrated in the Spanish Biobank Network), the for cancer- and aging dmCpGs in human and mouse, using Biobank of Galicia Sur Health Research Institute (IISGS), and the hg38 or mm10 genomes as reference. HOMER was run the Aragon Health Sciences Institute in the framework of the under default settings and with the -keepOverlappingBg pa- Biobank of Aragon. Specifically, they thank Laura Arregui rameter, using as background the filtered universe of CpG Egido (HUB-ICO-IDIBELL), Delia Recalde Frison (Biobank of sites analyzed by the RRBS for human and mouse, respec- Aragon) and Susana Teijeira Bautista, Vanesa Val Varela, and tively. Significant enrichments for known TFs were first fil- Olga Souto Rodrıguez (IISGS) for their contribution in the tered by FDR < 0.05, and subsequently only the top 50 results collection of the samples. This work was supported by the ordered by enrichment with respect to the background were Spanish Association Against Cancer (PROYE18061FERN to retained for each comparison (see supplementary table 8, M.F.F.), the Asturias Government (PCTI) cofunding 2018- Supplementary Material online, for full results). 2022/FEDER (IDI/2018/146 to M.F.F.), Fundacion General CSIC (0348_CIE_6_E to M.F.F.), and the Health Institute Liftover of Orthologous CpGs Carlos III (Plan Nacional de I þ DþI) cofunding FEDER Liftover of mm10 to hg38 coordinates was performed with (PI15/00892 and PI18/01527 to M.F.F and A.F.F.). They also UCSC’s liftOver tool (v377) (Hinrichs 2006) by using chain acknowledge support from the Ramon Areces Foundation files from https://hgdownload.soe.ucsc.edu/downloads.html (CIVP18A3891 to P.J.F.M.), the AECC (SIRTBIO to P.J.F.M.), last accessed April 29, 2021. Lifted CpG sites which did not the MICINN (SAF2017-85766-R to P.J.F.M.), a Ramon y Cajal reciprocally lift back from hg38 to the original mm10 coor- fellowship (MICINN, RYC-2017-22335 to P.J.F.M.), and the dinates were discarded ( 2%). European Commission ATTRACT project (777222 to A.P.C.). J.R.T. is supported by a Juan de la Cierva fellowship Conservation Analyses from the Spanish Ministry of Science and Innovation (FJCI- Base-resolution PhastCons tracks for 100-way (hg38) and 60- 2015-26965). R.F.P. and P.S.O. are supported by the Severo way (mm10) alignments (Siepel et al. 2005) were accessed via Ochoa program (BP17-114 and BP17-165, respectively). the GenomicScores R package (v1.1.10) (Puigdevall and R.G.U. is supported by the Centro de Investigacion Castelo 2018). Biomedica en Red de Enfermedades Raras (Health Institute Carlos III). L.V. was supported by the UAB Predoctoral training Gene Expression Analyses programme (PIF predoctoral fellowships). They also acknowl- Gene and isoform-level gene expression data in the form of edge support from the IMDEA Food Institute and IUOPA- RSEM counts and curated survival data were accessed via the ISPA-FINBA (the IUOPA is supported by the Obra Social Broad Institute Genomic Data Analysis Center (http://gdac. Cajastur-Liberbank, Spain). 3432 Aging and Cancer Epigenetic Signatures across Human and Mouse . doi:10.1093/molbev/msab112 MBE Author Contributions Carrasco-Garcia E, Moreno-Cugnon L, Garcia I, Borras C, Revuelta M, Izeta A, Lopez-Lluch G, de Pancorbo MM, Vergara I, Vina J, et al.. 2 9. R.F.P., J.R.T., A.F.F., and M.F.F. conceived, coordinated, and SOX2 expression diminishes with ageing in several tissues in supervised the study. R.F.P. and J.R.T. designed all aspects of mice and humans. Mech Ageing Dev. 177:30–36. 10.1016/ the research and contributed equally to this work. R.F.P. and j.mad.2018.03.008 29574045 J.R.T. collected the data, performed computational analyses, Cavalcante RG, Sartor MA. 2017. annotatr: genomic regions in context. Bioinformatics 33(15):2381–2383. and wrote the manuscript. A.F.F., P.S.O., V.L.M., and R.G.U. Ceccarelli M, Barthel FP, Malta TM, Sabedot TS, Salama SR, Murray BA, analyzed and interpreted the data. N.V.S. provided clinical Morozova O, Newton Y, Radenbaugh A, Pagnotta SM, et al. 2016. samples and biological data for human samples. A.P.C. and Molecular profiling reveals biologically discrete subsets and path- L.V. generated the mouse glioma models. M.B., P.J.F.M., and ways of progression in diffuse glioma. Cell 164(3):550–563. M.S. generated the mouse aging models. All authors revised, Chen F, Zhang Q, Deng X, Zhang X, Chen C, Lv D, Li Y, Li D, Zhang Y, Li P, et al. 2018. Conflicts of CpG density and DNA methylation are read, and approved the final manuscript. proximally and distally involved in gene regulation in human and mouse tissues. Epigenetics 13(7):721–741. Cheng Q, Huang C, Cao H, Lin J, Gong X, Li J, Chen Y, Tian Z, Fang Z, Data Availability Huang J. 2019. A novel prognostic signature of transcription factors for the prediction in patients with GBM. Front Genet. 10:906. The data underlying this article are available in the article and Chou S-J, Tole S. 2019. Lhx2, an evolutionarily conserved, multifunctional in its Supplementary Material online. Additionally, extended regulator of forebrain development. Brain Res. 1705:1–14. data sets are available in Zenodo, at https://dx.doi.org/10. Cole JJ, Robertson NA, Rather MI, Thomson JP, McBryan T, Sproul D, 5281/zenodo.4573676; last accessed April 29, 2021, principally Wang T, Brock C, Clark W, Ideker T, et al. 2017. Diverse interventions consisting of scripts used in the analyses, extended results, the that extend mouse lifespan suppress shared age-associated epige- netic changes at critical gene regulatory regions. Genome Biol. matrices of preprocessed methylation values and the lists of 18(1):58. orthologous CpG sites, the databases of histones, repetitive Conway JR, Lex A, Gehlenborg N. 2017. UpSetR: an R package for the DNA elements and chromatin states, information regarding visualization of intersecting sets and their properties. Bioinformatics the RNA-seq data analyzed. The sequencing files from the 33(18):2938–2940. RRBS experiments have been deposited in the European Day C-P, Merlino G, Van Dyke T. 2015. Preclinical mouse cancer models: a maze of opportunities and challenges. Cell 163(1):39–53. Nucleotide Archive (ENA) under the accession number Dennis DJ, Han S, Schuurmans C. 2019. bHLH transcription factors in PRJEB41460 (mouse data) and in the European Genome- neural development, disease, and reprogramming. Brain Res. Phenome Archive (EGA) under the accession number 1705:48–65. EGAS00001004851 (human data). Finally, an R Shiny app is Dmitrijeva M, Ossowski S, Serrano L, Schaefer MH. 2018. Tissue-specific available at https://epilabasturias.shinyapps.io/mbe_app/; last DNA methylation loss during ageing and carcinogenesis is linked to chromosome structure, replication timing and cell division rates. accessed April 29, 2021 which describes and facilitates access Nucleic Acids Res. 46(14):7022–7039. to the results of the differential methylation analyses. Du J, Johnson LM, Jacobsen SE, Patel DJ. 2015. DNA methylation path- ways and their crosstalk with histone methylation. Nat Rev Mol Cell Biol. 16(9):519–532. References Dutta S, Sengupta P. 2016. Men and mice: relating their ages. Life Sci. A lin A, Kormaksson M, Li S, Garrett-Bakelman FE, Figueroa ME, Melnick 152:244–248. A, Mason CE. 2012. methylKit: a comprehensive R package for the Edwards JR, O’Donnell AH, Rollins RA, Peckham HE, Lee C, Milekic MH, analysis of genome-wide DNA methylation profiles. Genome Biol. Chanrion B, Fu Y, Su T, Hibshoosh H, et al. 2010. Chromatin and 13(10):R87. sequence features that define the fine and gross structure of geno- Anders S, Huber W. 2010. Differential expression analysis for sequence mic methylation patterns. Genome Res. 20(7):972–980. count data. Genome Biol. 11(10):R106. Elango N, Yi SV. 2008. DNA methylation and structural and functional Aunan JR, Cho WC, Søreide K. 2017. The biology of aging and cancer: a bimodality of vertebrate promoters. Mol Biol Evol. 25(8):1602–1608. brief overview of shared and divergent molecular hallmarks. A&D Emerson RO, Thomas JH. 2009. Adaptive evolution in zinc finger tran- 8(5):628. scription factors. PLoS Genet. 5(1):e1000325. Baylin SB, Jones PA. 2016. Epigenetic determinants of cancer. Cold Spring Ernst J, Kellis M. 2010. Discovery and characterization of chromatin states Harb Perspect Biol. 8(9):a019505. for systematic annotation of the human genome. Nat Biotechnol. Beisel C, Paro R. 2011. Silencing chromatin: comparing modes and 28(8):817–825. mechanisms. Nat Rev Genet. 12(2):123–135. Ernst J, Kellis M. 2012. ChromHMM: automating chromatin-state dis- Blahnik KR, Dou L, Echipare L, Iyengar S, O’Geen H, Sanchez E, Zhao Y, covery and characterization. Nat Methods. 9(3):215–216. Marra MA, Hirst M, Costello JF, et al. 2011. Characterization of the Espada J, Esteller M. 2013. Mouse models in epigenetics: insights in de- contradictory chromatin signatures at the 30 exons of zinc finger velopment and disease. Brief Funct Genomics. 12(3):279–287. genes. PLoS One 6:e17121. Fernandez AF, Bayon GF, Urdinguio RG, Tora~ no EG, Garcıa MG, Carella Blewitt M, Whitelaw E. 2013. The use of mouse models to study epi- A, Petrus-Reurer S, Ferrero C, Martinez-Camblor P, Cubillo I, et al. genetics. Cold Spring Harb Perspect Biol. 5(11):a017939. 2015. H3K4me1 marks DNA regions hypomethylated during aging Bock C, Tomazou EM, Brinkman A, Müller F, Simmer F, Gu H, J€ager N, in human stem and differentiated cells. Genome Res. 25(1):27–40. Gnirke A, Stunnenberg HG, Meissner A. 2010. Genome-wide map- Ferrer-Font L, Villama~ nan L, Arias-Ramos N, Vilardell J, Plana M, Ruzzene ping of DNA methylation: a quantitative technology comparison. M, Pinna L, Itarte E, Ar us C, Candiota A. 2017. Targeting protein Nat Biotechnol. 28(10):1106–1114. kinase CK2: evaluating CX-4945 potential for GL261 glioblastoma Brennan CW, Verhaak RGW, McKenna A, Campos B, Noushmehr H, therapy in immunocompetent mice. Pharmaceuticals 10(4):24. Salama SR, Zheng S, Chakravarty D, Sanborn JZ, Berman SH, et al. Foronda M, Martınez P, Schoeftner S, Gomez-Lopez G, Schneider R, 2013. The somatic genomic landscape of glioblastoma. Cell Flores JM, Pisano DG, Blasco MA. 2014. Sox4 links tumor suppression 155(2):462–477. 3433 Perez et al. . doi:10.1093/molbev/msab112 MBE to accelerated aging in mice by modulating stem cell activation. Cell Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G, Rep. 8(2):487–500. Abecasis G, Durbin R, 1000 Genome Project Data Processing Gorkin DU, Barozzi I, Zhao Y, Zhang Y, Huang H, Lee AY, Li B, Subgroup. 2009. The Sequence Alignment/Map format and Chiou J, Wildberg A, Ding B, et al. 2020. An atlas of dynamic SAMtools. Bioinformatics 25(16):2078–2079. chromatin landscapes in mouse fetal development. Nature Liberzon A, Birger C, Thorvaldsdottir H, Ghandi M, Mesirov JP, Tamayo 583(7818):744–751. P. 2015. The Molecular Signatures Database (MSigDB) hallmark gene Greenberg MVC, Bourc’his D. 2019. The diverse roles of DNA methyla- set collection. Cell Syst. 1(6):417–425. tion in mammalian development and disease. Nat Rev Mol Cell Biol. Lister R, Mukamel EA, Nery JR, Urich M, Puddifoot CA, Johnson ND, 20(10):590–607. Lucero J, Huang Y, Dwork AJ, Schultz MD, et al. 2013. Global epi- Grüning B, Dale R, Sjödin A, Chapman BA, Rowe J, Tomkins-Tinch CH, genomic reconfiguration during mammalian brain development. Valieris R, Köster J. 2018. Bioconda: sustainable and comprehensive Science 341(6146):1237905–1237905. software distribution for the life sciences. Nat Methods. Liu XS, Wu H, Ji X, Stelzer Y, Wu X, Czauderna S, Shu J, Dadon D, Young 15(7):475–476. RA, Jaenisch R. 2016. Editing DNA methylation in the mammalian Hadad N, Masser DR, Blanco-Berdugo L, Stanford DR, Freeman WM. genome. Cell 167(1):233–247.e17. 2019. Early-life DNA methylation profiles are indicative of age-related Long HK, Sims D, Heger A, Blackledge NP, Kutter C, Wright ML, Grützner transcriptome changes. Epigenet Chromatin. 12(1):58. F, Odom DT, Patient R, Ponting CP, et al. 2013. Epigenetic conser- Hahn O, Grönke S, Stubbs TM, Ficz G, Hendrich O, Krueger F, Andrews S, vation at gene regulatory elements revealed by non-methylated Zhang Q, Wakelam MJ, Beyer A, et al. 2017. Dietary restriction DNA profiling in seven vertebrates. eLife 2:e00348. protects from age-associated DNA methylation and induces epige- Maegawa S, Gough SM, Watanabe-Okochi N, Lu Y, Zhang N, Castoro RJ, netic reprogramming of lipid metabolism. Genome Biol. 18(1):56. Estecio MRH, Jelinek J, Liang S, Kitamura T, et al. 2014. Age-related Han S, Liu Y, Cai SJ, Qian M, Ding J, Larion M, Gilbert MR, Yang C. 2020. epigenetic drift in the pathogenesis of MDS and AML. Genome Res. IDH mutation in glioma: molecular mechanisms and potential ther- 24(4):580–591. apeutic targets. Br J Cancer. 122(11):1580–1589. Maegawa S, Lu Y, Tahara T, Lee JT, Madzo J, Liang S, Jelinek J, Colman RJ, Hanna CW, Demond H, Kelsey G. 2018. Epigenetic regulation in devel- Issa J-PJ. 2017. Caloric restriction delays age-related methylation drift. opment: is the mouse a good model for the human? Hum Reprod Nat Commun. 8(1):539. Update. 24(5):556–576. Masser DR, Hadad N, Porter HL, Mangold CA, Unnikrishnan A, Ford Heinz S, Benner C, Spann N, Bertolino E, Lin YC, Laslo P, Cheng JX, Murre MM, Giles CB, Georgescu C, Dozmorov MG, Wren JD, et al. 2017. C, Singh H, Glass CK. 2010. Simple combinations of lineage- Sexually divergent DNA methylation patterns with hippocampal determining transcription factors prime cis-regulatory elements re- aging. Aging Cell 16(6):1342–1352. quired for macrophage and B cell identities. Mol Cell. 38(4):576–589. McKinney BC, Lin C-W, Rahman T, Oh H, Lewis DA, Tseng G, Sibille E. Hernando-Herraez I, Evano B, Stubbs T, Commere P-H, Jan Bonder M, 2019. DNA methylation in the human frontal cortex reveals a pu- Clark S, Andrews S, Tajbakhsh S, Reik W. 2019. Ageing affects DNA tative mechanism for age-by-disease interactions. Transl Psychiatry. methylation drift and transcriptional cell-to-cell variability in mouse 9(1):39. muscle stem cells. Nat Commun. 10(1):4361. Mendizabal I, Shi L, Keller TE, Konopka G, Preuss TM, Hsieh T-F, Hu E, Hernando-Herraez I, Garcia-Perez R, Sharp AJ, Marques-Bonet T. 2015. Zhang Z, Su B, Yi SV. 2016. Comparative methylome analyses iden- DNA methylation: insights into human evolution. PLoS Genet. tify epigenetic regulatory loci of human brain evolution. Mol Biol 11(12):e1005661. Evol. 33(11):2947–2959. Hernando-Herraez I, Heyn H, Fernandez-Callejo M, Vidal E, Fernandez- Michalak EM, Burr ML, Bannister AJ, Dawson MA. 2019. The roles of Bellon H, Prado-Martinez J, Sharp AJ, Esteller M, Marques-Bonet T. DNA, RNA and histone methylation in ageing and cancer. Nat Rev 2015. The interplay between DNA methylation and sequence Mol Cell Biol. 20(10):573–589. divergence in recent human evolution. Nucleic Acids Res. Nordman JC, Kabbani N. 2012. An interaction between a7 nicotinic 43(17):8204–8214. receptors and a G-protein pathway complex regulates neurite Hinrichs AS, Karolchik D, Baertsch R, Barber GP, Bejerano G, Clawson H, growth in neural cells. J Cell Sci. 125(22):5502–5513. Diekhans M, Furey TS, Harte RA, Hsu F, Hillman-Jackson J. 2006. The Oku Y, Huganir RL. 2013. AGAP3 and Arf6 regulate trafficking of AMPA UCSC Genome Browser Database: update 2006. Nucleic Acids Res. receptors and synaptic plasticity. J Neurosci. 33(31):12586–12598. 34(Database issue):D590–D598. Perez RF, Tejedor JR, Bayon GF, Fernandez AF, Fraga MF. 2018. Distinct Hirst TC, Vesterinen HM, Sena ES, Egan KJ, Macleod MR, Whittle IR. chromatin signatures of DNA hypomethylation in aging and cancer. 2013. Systematic review and meta-analysis of temozolomide in an- Aging Cell 17(3):e12744. imal models of glioma: was clinical efficacy predicted? Br J Cancer. Puigdevall P, Castelo R. 2018. GenomicScores: seamless access to 108(1):64–71. genomewide position-specific scores from R and Bioconductor. Huntley S, Baggott DM, Hamilton AT, Tran-Gyamfi M, Yang S, Kim J, Bioinformatics 34(18):3208–3210. Gordon L, Branscomb E, Stubbs L. 2006. A comprehensive catalog of Regnard C, Fesquet D, Janke C, Boucher D, Desbruyeres E, Koulakoff A, human KRAB-associated zinc finger genes: insights into the evolu- Insina C, Travo P, Edde B. 2003. Characterisation of PGs1, a subunit tionary history of a large family of transcriptional repressors. Genome of a protein complex co-purifying with tubulin polyglutamylase. J Res. 16(5):669–677. Cell Sci. 116(Pt 20):4181–4190. Karolchik D. 2004. The UCSC Table Browser data retrieval tool. Nucleic Roadmap Epigenomics Consortium, Kundaje A, Meuleman W, Ernst J, Acids Res. 32(90001):493D–496. Bilenky M, Yen A, Heravi-Moussavi A, Kheradpour P, Zhang Z, Wang Kessler NJ, Van Baak TE, Baker MS, Laritsky E, Coarfa C, Waterland RA. J, Ziller MJ, et al. 2015. Integrative analysis of 111 reference human 2016. CpG methylation differences between neurons and glia are epigenomes. Nature 518(7539):317–330. highly conserved from mouse to human. Hum Mol Genet. Ross JP, Rand KN, Molloy PL. 2010. Hypomethylation of repeated DNA 25(2):223–232. sequences in cancer. Epigenomics 2(2):245–269. Köppen M, Simske JS, Sims PA, Firestein BL, Hall DH, Radice AD, Rongo Schroeder DI, Jayashankar K, Douglas KC, Thirkill TL, York D, Dickinson C, Hardin JD. 2001. Cooperative regulation of AJM-1 controls junc- PJ, Williams LE, Samollow PB, Ross PJ, Bannasch DL, et al. 2015. Early tional integrity in Caenorhabditis elegans epithelia. Nat Cell Biol. developmental and evolutionary origins of gene body DNA meth- 3(11):983–991. ylation patterns in mammalian placentas. PLoS Genet. 11:e1005442. Krueger F, Andrews SR. 2011. Bismark: a flexible aligner and meth- Sen P, Shah PP, Nativio R, Berger SL. 2016. Epigenetic mechanisms of ylation caller for bisulfite-Seq applications. Bioinformatics longevity and aging. Cell 166(4):822–839. 27(11):1571–1572. 3434 Aging and Cancer Epigenetic Signatures across Human and Mouse . doi:10.1093/molbev/msab112 MBE Sheffield NC, Bock C. 2016. LOLA: enrichment analysis for genomic re- Sun D, Luo M, Jeong M, Rodriguez B, Xia Z, Hannah R, Wang H, Le T, gion sets and regulatory elements in R and Bioconductor. Faull KF, Chen R, et al. 2014. Epigenomic profiling of young and aged Bioinformatics 32(4):587–589. HSCs reveals concerted changes during aging that reinforce self-re- Shimizu D, Masuda T, Sato K, Tsuruda Y, Otsu H, Kuroda Y, Eguchi H, newal. Cell Stem Cell 14(5):673–688. Kodera Y, Mimori K. 2019. CRMP5-associated GTPase (CRAG) is a Therneau TM, Grambsch PM. 2000. Modeling survival data: extending candidate driver gene for colorectal cancer carcinogenesis. the Cox model. New York: Springer-Verlag. Anticancer Res. 39(1):99–106. Wang M, Zhao Y, Zhang B. 2015. Efficient test and visualization of multi- Siepel A, Bejerano G, Pedersen JS, Hinrichs AS, Hou M, Rosenbloom K, set intersections. Sci Rep. 5(1):16923. Clawson H, Spieth J, Hillier LW, Richards S, et al. 2005. Evolutionarily Wang T, Tsui B, Kreisberg JF, Robertson NA, Gross AM, Yu MK, Carter H, conserved elements in vertebrate, insect, worm, and yeast genomes. Brown-Borg HM, Adams PD, Ideker T. 2017. Epigenetic aging signa- Genome Res. 15(8):1034–1050. tures in mice livers are slowed by dwarfism, calorie restriction and Sizemore GM, Pitarresi JR, Balakrishnan S, Ostrowski MC. 2017. The ETS rapamycin treatment. Genome Biol. 18(1):57. family of oncogenic transcription factors in solid tumours. Nat Rev Wickham H. 2016. ggplot2: elegant graphics for data analysis. Springer Cancer. 17(6):337–351. International Publishing. ISBN 978-0-387-98140-6 Smit AF. 1996. The origin of interspersed repeats in the human genome. Yin T, Cook D, Lawrence M. 2012. ggbio: an R package for extending the Curr Opin Genet Dev. 6(6):743–748. grammar of graphics for genomic data. Genome Biol. 13(8):R77. Song H, Liu J, Wu X, Zhou Y, Chen X, Chen J, Deng K, Mao C, Huang S, Liu Yu M, Hazelton WD, Luebeck GE, Grady WM. 2020. Epigenetic aging: Z. 2019. LHX2 promotes malignancy and inhibits autophagy via more than just a clock when it comes to cancer. Cancer Res. mTOR in osteosarcoma and is negatively regulated by miR-129- 80(3):367–374. 5p. Aging 11(21):9794–9810. Zhou J, Sears RL, Xing X, Zhang B, Li D, Rockweiler NB, Jang HS, Srivastava R, Srivastava R, Ahn SH. 2016. The epigenetic path- Choudhary MNK, Lee HJ, Lowdon RF, et al. 2017. Tissue-specific ways to ribosomal DNA silencing. Microbiol Mol Biol Rev. DNA methylation is conserved across human, mouse, and rat, and 80(3):545–563. driven by primary sequence conservation. BMC Genomics 18(1):724. 3435